Figures and data

Allelic diversity in the Himba follows or somewhat exceeds linear predictions based on sample size from publicly available cohorts.
Allelic diversity for the 8 genotyped HLA genes was measured in the set of 102 unrelated Himba. African cohorts are labeled in orange, non-African cohorts in blue, and the Himba with a green diamond. Linear predictions of the number of unique alleles based on sample size were constructed from the available datasets, excluding the Himba. Ribbons indicate the 95% confidence interval for the linear prediction of the number of unique alleles based on sample size.

No effect of partnership types on genome-wide IBD sharing.
A) We consider a linear model where the intercept is the expected genome-wide IBD sharing based on a background distribution of unrelated female - male pairs from the population. We plot the effect size estimates of arranged partnerships and chosen partnerships on IBD sharing and find that the 95% confidence intervals of the effects contain zero, indicating there is no significant effect of arranged or chosen partnerships on IBD sharing. B) The distributions of the IBD sharing (cM) for arranged and chosen partnerships clearly overlap. Means of each distribution are noted with a dashed line.

No effect of partnership types on HLA similarity or average log odds of offspring HLA homozygosity.
We again consider linear models where the intercept is the expected outcome based on a background distribution of random, unrelated, opposite sex pairs from the population and control for genome-wide IBD sharing. We apply the model to the HLA similarity score (A) and to the average log odds of HLA homozygous offspring (C) across all 8 genes. We plot the effect size estimates of arranged partnerships and chosen partnerships on HLA similarity score (A) and average log odds of offspring homozygosity (C) and find the 95% confidence intervals contain zero, indicating no significant effect of arranged or chosen partnerships on HLA similarity or average log odds of offspring homozygosity. The distributions of HLA similarity scores (C) and average log odds of offspring HLA homozygosity (D) clearly overlap between arranged and chosen partnerships. Means of each distribution are noted with a dashed line.

Elevated IBD sharing at the HLA region is consistent with recent, fluctuating positive selection.
Panels A and B depict IBD sharing rates across chromosome 6 (rates were averaged within 1000 base pair bins). The dashed line depicts the genome-wide median IBD sharing rate, excluding chromosome 6. Panel A depicts IBD sharing rates for the pairs that make up the background distribution for the partnership analyses (random heterosexual pairings, with relevant age gaps, related at no closer than 3rd degree; n=352 individuals). Panel B depicts the IBD sharing rates for the unrelated individuals included in the allele frequency estimations (related at no closer than 4th degree) with the common haplotypes shown in Figure S9 (n=71 individuals). The HLA region is designated in grey. The pink shaded region depicts +2 and -2 standard deviations from the genome-wide (excluding chromosome 6) median IBD sharing rate, 3.1% and 2.8% for panels A and B, respectively.